>Your entire line of reasoning is based on the assumption that risk of fire is unavoidable and such we don't want to exert pressure on them to reduce that risk

>Your entire line of reasoning is based on the assumption that risk of fire is unavoidable and such we don't want to exert pressure on them to reduce that risk any further than regulation requires. Not quite. I think some risk of fire is unavoidable but I'm not saying they should do only what the regulatory minimum is and no more. If we assume that there is a relationship between the amount of "effort" (whether that is money, management insight, or whatever) they put in and the fire risk then we want them to be incentivised through most of the range of that function. If they are good at preventing fires, investors and management should benefit, if they are bad then they should pay penalties. The point I'm making if that single events can cause 10s of billions of dollars of damage, and if the trigger for that event is something which can be reduced but not eliminated, then the incentive structure ends up being much more like a lottery ticket than a smooth relationship that rewards diligent management. If they were instead fined a substantial amount (say $100m) every time one of their assets caused any kind of fire in a risky area without being liable for all the damages (which is not within their control) then I think that would drive a much more effective set of incentives. It's fundamentally a contrast between a light-touch regulatory mechanism with uncapped tort liability + insurers requiring certain behaviour and a heavier regulatory incentive where a regulator requires that behaviour but with the liability of the consequences held elsewhere. To be fair, in many cases the US uses the former and it may just be my European utility regulation perspective that makes me think it is odd since we typically use mechanisms of the second sort. My intuition says that when the cost of the consequences has a power-law type distribution (most fires cost very little but a few cost billions), it is more efficient to use the latter model rather than rely on insurance models but that's admittedly not based on any detailed analysis.